Study of 3D printing in optimal design of structures and materials
Document typeMaster thesis
Rights accessRestricted access - author's decision
Topology optimization (TO) is a promising numerical technique for designing optimal engineering structures in many industrial applications. It is expected that it might become an unavoidable engineering tool for many new rising technologies such as the additive manufacturing or metal 3D printing. Since these numerical methods and tools are dispersed in the literature, the purpose of this study is to benchmark different optimization solvers when applied to various finite element based structural topology optimization problems, for both structural and material design. Different optimization solvers including the Method of Moving Asymptotes (MMA), the interior point solver IPOPT and Augmented Lagrangian scheme in combination with the SLERP (when using a Level Set function) and the Projected Gradient (when using density-like variables) are compared. To this aim, the problems are solved by exploring on a new material interpolation scheme, SIMP-ALL, which is based on the topological derivative combined with a density filter. Furthermore, this new interpolation scheme is compared as well to the widely accepted SIMP interpolation scheme. Different examples of optimum structures are presented for many cases including macro-scale, micro-scale, compliant mechanism and inverse problems. In addition, a new incremental optimization technique is presented as a resource for better control of the optimization process and for improving convergence in very numerically difficult cases similarly to what is done in plasticity problems. Finally some of the obtained designs are manufactured using 3D printing techniques, in order to prove the manufacturability of the designs for future industrial applications. In this regard, special mention of perimeter control in topology optimization is made since it is a key tool for ensuring manufacturability of the designs, reducing “grey” areas and checkerboard patterns and improving convergence in determined problems.